Transparent huge pages merged in Linux 2.6.38

Discussion of chess software programming and technical issues.

Moderator: Ras

UncombedCoconut
Posts: 319
Joined: Fri Dec 18, 2009 11:40 am
Location: Naperville, IL

Transparent huge pages merged in Linux 2.6.38

Post by UncombedCoconut »

This news is almost 2 months old, but I just noticed it. It seems very relevant for CC programs. If I'm not missing something: It now makes sense for engines always to request huge pages on Linux. More importantly, it makes sense for testers and developers to disable the feature via sysfs before running test matches, to avoid situations where the kernel only gives one player huge pages in a game.

On this subject, are there many other non-obvious pitfalls for testers? (Maybe this is a good time to start a CPW page on the subject.)
rbarreira
Posts: 900
Joined: Tue Apr 27, 2010 3:48 pm

Re: Transparent huge pages merged in Linux 2.6.38

Post by rbarreira »

Very interesting, thanks for posting this.
Gerd Isenberg
Posts: 2251
Joined: Wed Mar 08, 2006 8:47 pm
Location: Hattingen, Germany

Re: Transparent huge pages merged in Linux 2.6.38

Post by Gerd Isenberg »

UncombedCoconut wrote:This news is almost 2 months old, but I just noticed it. It seems very relevant for CC programs. If I'm not missing something: It now makes sense for engines always to request huge pages on Linux. More importantly, it makes sense for testers and developers to disable the feature via sysfs before running test matches, to avoid situations where the kernel only gives one player huge pages in a game.

On this subject, are there many other non-obvious pitfalls for testers? (Maybe this is a good time to start a CPW page on the subject.)
Thanks, so far I put some links under Memory - Huge Pages. Clearly deserves an own page if more material e.g. code snippets for Linux/Windows is added.
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: Transparent huge pages merged in Linux 2.6.38

Post by bob »

Gerd Isenberg wrote:
UncombedCoconut wrote:This news is almost 2 months old, but I just noticed it. It seems very relevant for CC programs. If I'm not missing something: It now makes sense for engines always to request huge pages on Linux. More importantly, it makes sense for testers and developers to disable the feature via sysfs before running test matches, to avoid situations where the kernel only gives one player huge pages in a game.

On this subject, are there many other non-obvious pitfalls for testers? (Maybe this is a good time to start a CPW page on the subject.)
Thanks, so far I put some links under Memory - Huge Pages. Clearly deserves an own page if more material e.g. code snippets for Linux/Windows is added.
The nice thing about the Linux kernel changes is that you don't have to do anything. You can suggest that huge pages will be good. Or you can suggest that normal pages are better. If you suggest nothing, it will figure it out although it might take a few seconds or even minutes to get it all worked out.

I had followed some of the discussions, and was hoping some of this would make it into the kernel. The linux approach makes perfect sense, since the automatic way is always the best way, as opposed to users having to do something unique in their code...

I'm going to install this on my laptop pretty quickly to see how it behaves...